No CrossRef data available.
Article contents
Representation from Bottom and Top
Published online by Cambridge University Press: 01 January 2020
Extract
I would like to nominate one more principle for initial inclusion in the science of teleonomy. This principle is that the nature of the stimuli that initiate and regulate a response may be no indication of the function of the response.
George Williams could not have anticipated the special relevance his principle has for contemporary analyses of representational content. In particular, his principle provides both a concise statement of where a currently popular strategy for naturalizing representational content has gone wrong and a positive suggestion for how we should right this wrong. I characterize the kind of naturalistic analysis of representation I have in mind as bottom-up because it seeks to build representation up from a non-intentional, and hence naturalistically unimpeachable, correlation relation. Many authors have suggested such an approach to naturalizing intentionality, but for clarity and completeness perhaps Fred Dretske's Explaining Behavior: Reasons in a World of Causes ought to be construed as the exemplary source.
- Type
- Research Article
- Information
- Copyright
- Copyright © The Authors 1996
References
2 Williams, George Adaptation and Natural Selection (Princeton, NJ: Princeton University Press 1966), 269Google Scholar
3 See also Stampe, Dennis ‘Toward a Causal Theory of Linguistic Representation,’ in French, P. Uehling, T. and Wettstein, H. eds., Midwest Studies in Philosophy II (Minneapolis: University of Minnesota Press 1977) 42–63Google Scholar; Fodor, Jerry Psychosemantics: The Problem of Meaning in the Philosophy of Mind (Cambridge, MA: The MIT Press 1987)Google Scholar; Matthen, Mohan ‘Biological Functions and Perceptual Content,’ Journal of Philosophy 85 (1988) 5–27CrossRefGoogle Scholar.
4 Dretske, Fred Explaining Behavior: Reasons in the World of Causes (Cambridge, MA: The MIT Press 1988)Google Scholar
5 Pittendrigh, Colin ‘Adaptation, Natural Selection, and Behavior,’ in Roe, A. and Simpson, G. eds., Behavior and Evolution (New Haven, CT: Yale University Press 1958) 390–416Google Scholar
6 Genetic mutation and recombination will produce variance in the success of individuals’ responses to particular selection pressures.
7 Pittendrigh's discussion of the mechanisms by which fruit flies prevent desiccation turns out to be greatly oversimplified. More recent work, e.g., Seiger, Marvin and Kertesz, Joseph ‘The Effect of Relative Humidity on Photoresponse in Sympatric Species of Drosophila: Short Term Exposure to Desiccating Environments,’ Journal of Insect Physiology 33 (1987) 477–80Google Scholarand Seiger, Marvin and Woodruff, David ‘The Effect of Relative Humidity on Photoresponse in Sympatric Species of Drosophila: Long-term Exposure to Desiccating Environments,’ Journal of Insect Physiology 33 (1987) 529–32CrossRefGoogle Scholar, suggests that fruit flies do not simply seek out the most humid environments. When exposed to desiccating environments in the short term they are able to maintain their water balance by means of transpiration and so their behavior may not exhibit any photosensitivity. In the long term they will display photonegative behavior in environments with extremely low relative humidity (e.g. 5%) but photopositive behavior in environments with higher relative humidity (e.g. 50%). Photopositive behavior may expose the fruit fly to a dryer environment but can enable the fly to find a source of water with which to rehydrate itself. In the following I intend to ignore these complications. However, I recommend the reader keep these complications in mind, because they only reinforce the point I shall make later about the trade-offs that must figure into the selection of indicators.
8 For example, in Explaining Behavior, Dretske says, ‘electrically operated fuel gauges indicate not only the amount of fuel left in the tank but also the amount of electrical current flowing in the wires connecting the gauge to the tank, the amount of torque on the armature to which the pointer is affixed, and the magnitude of the magnetic field surrounding this armature. Given the way these gauges operate, they cannot indicate (i.e. have their behavior depend on) the amount of fuel in the tank without indicating (exhibiting at least the same degree of dependency on) these related conditions.’ Despite the number of conditions an electric fuel gauge indicates, it represents only that condition it has the function to indicate: the amount of fuel in the tank. (59)
9 Sometimes a token of the type indicator of F (when broken, or removed from the environment in which it evolved), will not indicate F. However, it may still retain the function to indicate F, because it is still the indication of F that explains why this kind of indicator has been given the control duties it now has. When a token of the kind indicator of F ceases to be an indicator of F, it will misrepresent those conditions it now does indicate as F. For more on misrepresentation see Dretske, Explaining Behavior, 64–70Google Scholar and his ‘Misrepresentation,’ in Bogdan, R. ed., Belief Form, Content, and Function (Oxford: Clarendon Press 1986) 17–36Google Scholar.
10 Others have sought to refine Dretske's indication relation in various ways (see, e.g., Godfrey-Smith, Peter ‘Signal, Decision, Action,’ Journal of Philosophy 88 [1991] 709–22CrossRefGoogle Scholar; and Malcolm Forster, ‘How Neural Networks Develop Meaningful Representations Through Learning,’ [unpublished manuscript]) and I do not claim that the following argument is applicable to every understanding of indication. However, I will argue that it is not correlation that lies at the heart of representation.
11 In Explaining Behavior Dretske distinguishes between ‘mere’ representations and beliefs. When an indicator of F acquires the function to indicate F because of its selection history, it represents F (see 94). But, Dretske claims, selection cannot provide an account of the control duties of indicators within individual organisms (92). Evolution, he thinks, can't explain why this particular organism is wired the way it is. Rather, this particular organism is wired the way it is because of the genes it carries. So, Dretske distinguishes between representations and beliefs. When an indicator of G acquires the function to indicate G within the lifetime of an individual, it comes to be the belief that G. Dretske appeals to associative learning to explain the recruitment of indicators within single life spans.
12 In addition to Godfrey-Smith and Sober, Cummins, Robert Meaning and Mental Representation (Cambridge, MA: The MIT Press 1989), 162Google Scholar; and Dennett, Daniel ‘Evolution, Error, and Intentionality,’ in The Intentional Stance (Cambridge, MA: The MIT Press 1987) 304Google Scholar raise this objection. Godfrey-Smith, however, develops the objection most effectively, and it is his paper that brought to my attention the passages in Cummins and Dennett.
13 Godfrey-Smith, Peter ‘Indication and Adaptation,’ Synthese 92 (1992) 283–312CrossRefGoogle Scholar
14 Godfrey-Smith and I differ on our readings of indication. Godfrey-Smith assumes that indication can be a ‘more or less’ reliable relation. However, I believe that Dretske is quite explicit that C indicates F if and only if state C is present only if F obtains. As we will see, it is because Dretske has this strict definition of indication in mind that he is forced to introduce in addition to indication a weaker notion of correlation. Godfrey-Smith's objection to Dretske, however, can be made with either understanding of indication. And, to keep my analysis consistent, I will reframe Godfrey-Smith's objections using my characterization of indication.
15 Millikan, Ruth Language, Thought, and Other Biological Categories (Cambridge, MA: The MIT Press 1984)Google Scholar; ‘Biosemantics,’ Journal of Philosophy 86 (1989) 281-97; ‘Compare and Contrast Dretske, Fodor, and Millikan on Teleosemantics,’ Philosophical Topics 18 (1990) 151-61; ‘On Mentalese Orthography,’ in Dahlboom, B. ed., Dennett and His Critics (Oxford: Basil Blackwell 1993) 97–123Google Scholar
16 In Explaining Behavior, Dretske summarizes the Sober point like this: ‘for selection to take place all that is needed is for the triggering state to be better correlated with the appropriate season than are the corresponding states in competing plants. A state need not be reliably correlated with spring — hence need not indicate the arrival of spring — in order to be correlated sufficiently well with the arrival of spring to confer on its possessor a competitive advantage.’ That Sober's and Godfrey-Smith's observations are connected is fairly obvious. Both point out that selection will not always favor an indicator of what, following Pittendrigh, we may call the adaptively significant condition. Depending upon the costs and benefits involved, selection might favor a tree with a mild-weather indicator to a tree with a spring indicator. It is not, then, because trees have spring indicators that trees blossom in the spring, but because trees have indicators of something else that is correlated, correlated sufficiently well to be favorable, with spring. I do not wish to debate exactly how similar Sober's and Godfrey-Smith's points are. It suffices for what follows that Dretske is aware that selection will not always favor an indicator of an adaptively significant condition. (90)
17 Ibid., 102. Dretske goes on to say, with Sober's approval, that the better the correlation between the indicated condition and F, the better the solution. I think Godfrey-Smith successfully challenges this claim when he notes that cost-benefit considerations might still make more favorable an indicator of a condition less reliably correlated with F. I leave it to the reader's imagination to make this claim plausible.
18 Dennett, in ‘Evolution, Error, and Intentionality,’ has argued that there is an indeterminacy involved with function ascriptions and, accordingly, he might object that the kind of top down analysis of content I'm advocating here would confront similar problems of indeterminacy. However, there is a fact of the matter about which environmental conditions posed adaptive problems to members of a species. It is with this fact that the determinacy of functions and the content derived from these functions is anchored.
19 Shapiro, Lawrence ‘Content, Kinds, and Individualism in Marr's Theory of Vision,’ The Philosophical Review 102 (1993) 489–513Google Scholar
20 Dretske denies that he assigns the wrong content to the states of the stickleback's detector (personal communication). He claims that it is implausible that sticklebacks are capable of representing things as male sticklebacks. However, the sense of representation Dretske here employs, viz. representation as, is more sophisticated than the sense I have in mind and, I think, than is needed for much of cognitive psychology (e.g. the theory of vision developed in Marr, David Vision [San Francisco, CA: Freeman 1982]Google Scholar). For present purposes, I intend representation in its intuitive ‘stands for’ sense (for more on this sense of representation as it appears in Marr's theory, see Hatfield, Gary ‘Representation in Perception and Cognition: Connectionist Affordances,’ in Ramsey, W. Stich, S. and Rumelhart, D. eds., Philosophy and Connectionist Theory [Hillsdale, NJ: Lawrence Erlbaum Associates 1991] 163–95)Google Scholar. Hence, I do not have the difficulty Dretske does in conceiving that sticklebacks have states that represent male sticklebacks. Indeed, I find this claim more plausible than Dretske's denial that a stickleback, when responding to a fire truck, is misrepresenting.
21 Not only is this a tacit alternative, but it is one that Dretske would not, I think, want credited to him.
22 For an extended discussion of this distinction between conceptions of naturalization see Hatfield, Gary The Natural and the Normative: Theories of Spatial Perception from Kant to Helmholtz (Cambridge, MA: The MIT Press)Google Scholar. See also Epstein, William and Hatfield, Gary ‘Gestalt Psychology and the Philosophy of Mind,’ Philosophical Psychology 7 (1994) 163–81Google Scholar. In Hatfield's vocabulary, metaphysical naturalism (in philosophy of mind) is the view that mental states are natural only if they can be reduced to physiology and, subsequently, physics. Methodological naturalism, on the other hand, is the position that mental states are natural if their study is amenable to the investigatory standards of a natural science. For a more thorough criticism of the view of naturalism I call Lego Naturalism see my ‘The Nature of Nature: Rethinking Naturalistic Theories of Thought’ (unpublished manuscript).